Iteration
Итерация
Speed and scale Holds Depth 1 · Shovel
short cycle between improvements
Moats it grows into
- #23 Speed
Every cycle is a run of the loop that Speed measures: hypothesis, release, data. Run it more often than a rival and the learning rate is the edge, for as long as the cadence holds.
Depth 1; a real hold only when it compounds into #3 or #19.
Summary
Iteration is a short loop of change, release and measurement, run more often than a rival runs theirs. Its edge is a learning rate: five cycles of hypothesis and data in the time it takes someone else to ship one version. It holds, but at depth 1 — the tempo itself is the moat called Speed, a shovel, and it only deepens when the loop leaves something behind.
How it works
The mechanism is compounding correction. Each cycle produces a change, a release and a measurement; the measurement decides the next change. A team on a weekly cycle makes fifty corrections a year against a rival's four, and every correction is made on fresher evidence. The gap is not in any single release but in the accumulated aim: after a year the fast team has tested hypotheses the slow one has not had time to state.
What makes it work is the cost of a cycle, not the courage to ship. Cheap deploys, fast rollback, real usage data within hours and a habit of writing down what each release was supposed to move. Where a cycle costs a rocket, the loop is still the same: fly, break, measure, change the next vehicle. Where a cycle is a database migration nobody dares to run, the loop stalls and the tempo becomes a slide deck.
Why it holds — or only takes
Iteration holds because tempo is itself a moat: Speed, #23, a compounding advantage in learning rate. But Speed is a shovel — a rival matches it within months by adopting the same cadence, and AI tooling is handing that cadence to everyone. So the hold is real and shallow. The loop deepens only when each cycle deposits something a fast rival cannot deposit as quickly: usage data that becomes Proprietary data, or the thousands of small practices that become Process power. Tempo without a deposit is a lead, not a wall.
Do you have it
- A change you decided on Monday is in front of real users by Friday, with a number attached.
- Rollback is a routine command, not a meeting, so shipping a wrong guess costs a day.
- Your changelog for the last quarter lists more shipped experiments than your rival's for the year.
- Each release has a written expected outcome, and someone checks it against the actual one.
How to dig it
- Cut the cost of one cycle first: deploys under an hour, feature flags, one-command rollback.
- Instrument before you ship; a release without a measurement is a guess, not an iteration.
- Keep a log of hypothesis, result and decision; that log is the deposit the tempo leaves.
- Route what the loop learns into a dataset or a written practice, so Speed feeds a deeper moat.
How it erodes
- A rival adopts the same cadence; the tooling that made you fast is now a default in every stack.
- The product gets big and every cycle drags a migration; the weekly loop slips to quarterly.
- Users tire of constant change and start asking for a stable version; tempo turns into churn.
- A frightened incumbent enters burst mode and out-ships you for a quarter with ten times the hands.
Solo, without capital
This is the most solo-friendly hold in the set: Speed is reachable alone, needs no capital and cannot be rented, so a single builder with a small user base can out-cycle any team by pure cadence. What you cannot do is stay there. The tempo lead lasts months, and the deeper moats it feeds — a proprietary dataset, process power — need years or headcount. Use the year of tempo to accumulate the data.
What the matrix says of its moats
- #23 Speed Solo ✓ solo-reachable Capital ○ low AI ↓ under attack Rent ✗ not for sale
In nature, in business
- Nature
- —
- Business
- SpaceX
Cases
- SpaceX — Flies test vehicles it expects to lose, reads the failure, and changes the next one; the Starship campaign is the loop in the open.
- Toyota — Runs kaizen on the line: any worker can stop production, the cause is fixed in the process, and the next shift runs the improved version.
- Duolingo — Ships most product changes as A/B experiments and keeps only the variant that moves retention, so the app is rebuilt in small measured steps.
Combines with
- Second-mover — Copying the first mover buys the start; out-cycling them afterwards is what keeps the copy ahead of the original.
- Antifragility — A product that gets stronger under stress needs a loop that turns every failure into the next change; iteration is that loop.
- First-mover — Being first buys a few months; a faster cycle than the followers is the only way to spend them well.
Fights with
- Craftsmanship — Craftsmanship spends more time on each unit; a weekly cycle spends less, and one of the two has to give.
- Predictability — Users who chose you for a consistent experience pay for every change you ship; constant iteration taxes the very thing they bought.